On the distribution of dorsals in complex and simple onsets in child German, Dutch and English

The aim of this paper is to argue that there is substantive variation between child language and adult language, contrary to what is often assumed in acquisition theories.(1) Our data show that some children have a distribution of dorsals-allowed in clusters, but not in singletons-that is typologically unattested. The grammar needed to describe the language of the children is never needed to describe any adult language. In other words, there is appears to be variation in the substance of child and adult grammars.

In Optimality Theory (Prince and Smolensky, 1993/2004), a grammar is a hierarchical ranking of a set of universal constraints.(2) A stage in child language is a ranking of these constraints and should correspond to an adult language (Levelt and van de Vijver, 2004). We call this kind of variation between child grammars and adults grammars unsubstantive variation. An example of such unsubstantive variation comes from the acquisition of syllable types. Children acquiring Dutch have to learn that syllables with complex onsets and complex codas are allowed. The grammar of the children goes through a stage in which only simple onsets and codas are allowed. This stage can be described with the same constraints-markedness constraints against complex syllable margins outrank faithfulness constraints-that are necessary to describe the final stage-faithfulness constraints now outrank markedness constraints. Moreover, the stage in which only simple margins are allowed corresponds to the grammar of Thargari (Levelt and van de Vijver, 2004).

A substantive difference between child language and adult language, on the other hand, is found when child language has properties not found in any adult language. No adult language requires codas and, as a consequence, there is no constraint or constraint ranking that would cause epenthesis of consonants in case a syllable ends in a vowel. If a child language would require codas and we were to set up a constraint or a ranking that would ensure epenthesis of codas in CV syllables, there would be no typological justification. With respect to segments, there are no adlt languages in which segments that are allowed in complex onsets are not alowed in simple onsets. Yet, this is what we are faced with in the data discussed in this paper. We will show that it is possible to account for the data with positional markedness (Zoll, 2004). Even though the distribution can be explained elegantly in this way, a problem remains: there are no adult languages in which dorsals are distributed in the same way.